## Abstract

Increasing prevalence of overweight and obesity motivated this prospective examination of gestational diabetes mellitus in relation to self-reported adult height, weight, and weight fluctuation. Gestational diabetes was assessed by use of medical records in 1,644 women enrolled in Seattle and Tacoma, Washington, between 1996 and 2002. After adjustment, risk was inversely related to height and directly related to pregravid body mass index (ptrend < 0.001). The relation with body mass index at age 18 years was J shaped, with higher risk among lean women (adjusted relative risk (RR) = 1.79, 95% confidence interval (CI): 1.01, 2.84) and obese women (RR = 4.53, 95% CI: 1.25, 16.43) versus normal-weight women. Weight gain between age 18 years and the study pregnancy was associated with increased risk independently of body mass index at 18 years and other confounders (≥10-kg gain vs. <2.5-kg change: RR = 3.43, 95% CI: 1.60, 7.37). Weight cycling (loss and regain of ≥6.8 kg) was not associated after adjustment for body mass index at 18 years and adult weight change (≥3 vs. zero cycles: RR = 1.23, 95% CI: 0.56, 2.73). Cycling was nonsignificantly related among women who gained 10 kg or more during adulthood (≥3 vs. zero cycles: RR = 2.04, 95% CI: 0.83, 5.02). Efforts to prevent obesity and weight gain among young women may reduce gestational diabetes risk.

Gestational diabetes mellitus is the onset or first recognition of glucose intolerance during pregnancy (1). This condition affects 4–7 percent of pregnancies and is associated with adverse fetal outcomes such as macrosomia, jaundice, and birth trauma (1–3). Women with gestational diabetes are more likely than normoglycemic women to experience it again during a subsequent pregnancy, to experience other pregnancy complications such as preeclampsia, and to develop overt diabetes after pregnancy (4–6).

Several studies have shown that gestational diabetes is more common among overweight and obese women than among lean women (7–9). The current high prevalence of overweight and obesity among adult women in the United States (10) and the increasing prevalence worldwide (11) motivate detailed characterization of the aspects of weight and weight gain that may contribute to gestational diabetes risk. Using data from a prospective cohort study of pregnancy, we examined the association between gestational diabetes and measures of adult height, weight, and weight fluctuation. We sought to contribute to the small body of literature examining adult weight change as a potential risk factor for the disorder. Additionally, published studies of the relation between weight cycling and type two diabetes inspired us to examine weight cycling as a risk factor for gestational diabetes (12, 13).

## MATERIALS AND METHODS

### Study design and population

This prospective analysis is based on data collected from the Omega Study, an ongoing cohort study designed to examine risk factors of preeclampsia and gestational diabetes. Participants attended prenatal care clinics affiliated with the Swedish Medical Center and Tacoma General Hospital in Seattle and Tacoma, Washington, respectively. Eligible women were those who started prenatal care before 20 weeks of gestation, were 18 years of age or older, could speak and read English, and planned to carry the pregnancy to term and to deliver at either research hospital. During the first 18 months of the study, nulliparous women were preferentially invited to participate. A higher proportion of parous women were invited as the study progressed.

Trained research personnel interviewed participants using a structured questionnaire to gather information on sociodemographic and anthropomorphic characteristics (including height and weight characteristics, described below), lifestyle habits, and medical and reproductive histories. After delivery, participants' hospital labor and delivery medical records and clinic records were abstracted for information on pregnancy, delivery, and infants. The procedures used in this study were approved by the institutional review boards of the Swedish Medical Center and Tacoma General Hospital. All participants provided written informed consent.

### Analytical population

The analytical population comprises women who enrolled between September 1996 and December 2002. During this time, 2,000 (84 percent) of 2,381 eligible women who were approached for enrollment agreed to participate. For this analysis, we excluded 121 women with unknown delivery outcome or missing medical records. We also excluded 49 women who experienced an abortion or fetal death before 28 weeks of gestation, 21 with pregestational diabetes, 101 with chronic hypertension, and 65 with missing weight or height information. This analysis is based on information from 1,644 women.

### Gestational diabetes definition

We used the National Diabetes Data Group diagnostic criteria for gestational diabetes, which were current at the start of the study (14). Women were routinely screened for gestational diabetes between 24 and 28 weeks of gestation by use of a 50-g, 1-hour oral glucose tolerance test. Those with a posttest plasma glucose level of more than 140 mg/ml participated in the diagnostic 100-g, 3-hour oral glucose tolerance test. Women were classified as having gestational diabetes if two or more 3-hour test concentrations exceeded the following cutpoints: fasting, ≥105 mg/dl; 1 hour, ≥190 mg/dl; 2 hours, ≥165 mg/dl; 3 hours, ≥145 mg/dl (14).

### Weight and height characteristics

Interviewers asked each woman to report her height without shoes, her recalled weight at age 18 years, and her weight immediately before the study pregnancy (pregravid weight). We used pregravid body mass index to estimate adiposity. We categorized it according to Institute of Medicine cutpoints as <19.8 (lean), 19.8–26.0 (normal), 26.1–29.0 (overweight), and >29.0 kg/m2 (obese) (15). We categorized height according to approximate quartiles: ≤160, 161–165, 166–170, and >170 cm. We calculated the net change in weight from age 18 years to just before pregnancy and categorized it as follows: loss of 2.5 kg or more, net change of fewer than 2.5 kg, 2.6–4.9 kg gain, 5.0–9.9 kg gain, and gain of 10 kg or more. These categorizations were chosen a priori and were similar to those used in studies of weight change in relation to chronic diseases in older women (16, 17). We also analyzed adult weight gain as a continuous variable in 5-kg increments. Interviewers ascertained weight cycling by asking, “How many times have you lost 15 pounds (6.8 kg) or more and then later gained all the weight back (not counting when you were pregnant)?” This 15-pound cutpoint was arbitrary and chosen during questionnaire design. We classified women according to history of weight cycling (yes/no) and number of cycles (zero, one, two, and ≥3).

### Statistical analysis

We examined frequency distributions of participants' characteristics according to categories of adult net weight change. We used multivariable generalized linear models with log link functions to estimate adjusted relative risks and 95 percent confidence intervals (18). We tested for linear trend across categories of height and weight characteristics by modeling them as grouped linear variables. Two-tailed p values of less than 0.05 were considered statistically significant. We repeated all analyses after excluding 27 women who reported having gestational diabetes in a prior pregnancy. This exclusion did not meaningfully change the results (data not shown). Statistical analyses were performed using STATA, version 9.0, software (StataCorp LP, College Station, Texas).

We evaluated confounding for each exposure separately. The characteristics listed in table 1, as well as height and weight characteristics when appropriate, were evaluated as potential confounders. We entered each covariate into a regression model relating the exposure to gestational diabetes risk and compared models with and without adjustment for the covariate. We included in the final model covariates that altered unadjusted relative risks by 10 percent or more. For each covariate, we chose the coding specification that achieved the greatest control of confounding.

TABLE 1.

Characteristics of Omega Study participants (n = 1,644), by percent distributions overall and according to adult net weight change, Seattle and Tacoma, Washington, 1996–2002

 Characteristic Entire population (n = 1,644) Adult net weight change (kg) ≤−2.5 (n = 120) −2.4 to 2.5 (n = 389) 2.6 to 4.9 (n = 257) 5.0 to 9.9 (n = 401) ≥10.0 (n = 477) Age (years) <20 0.7 0.7 2.3 0.4 0.0 0.2 20–34 70.6 77.5 74.4 71.2 71.1 64.8 35–39 23.4 20.8 19.8 23.0 23.9 26.8 ≥40 5.3 0.8 3.3 5.4 5.0 8.2 Race/ethnicity White 84.8 82.5 84.6 88.7 85.5 83.0 African American 1.9 0.8 2.3 0.0 2.0 2.7 Hispanic 2.9 3.3 2.3 2.0 3.2 3.4 Asian 7.4 10.0 9.5 6.2 6.5 6.5 Other 2.8 3.3 1.3 3.1 2.5 4.0 Missing 0.2 0.0 0.0 0.0 0.3 0.4 Nulliparous 67.7 65.0 67.9 71.2 70.6 63.9 Employed 82.6 80.0 81.0 84.4 83.8 82.6 ≤12 years of education 4.5 4.2 4.9 3.9 4.5 4.6 Household income ($) <30,000 4.3 5.8 6.4 2.3 3.5 4.0 30,001–69,999 24.2 24.2 18.0 22.2 24.7 29.8 ≥70,000 68.9 65.8 72.2 74.4 69.3 64.6 Missing 2.7 4.2 3.4 3.1 2.5 1.7 Regular physical activity in the year before pregnancy 91.9 91.7 92.3 91.4 94.3 89.7 Smoking history Never 71.9 70.8 75.1 77.2 71.0 67.4 Before pregnancy 21.7 20.0 19.2 18.4 23.7 24.0 At beginning of pregnancy 6.4 9.2 19.2 18.4 23.7 24.0 Prenatal vitamin use 97.4 97.5 97.2 98.4 98.3 96.2 First-degree familial diabetes 14.1 9.2 12.1 12.8 14.0 17.6 Pregravid body mass index (kg/m2) <19.8 (lean) 18.1 43.2 39.3 21.4 8.0 1.1 19.8–26.0 (normal) 65.6 54.2 57.6 75.9 86.4 51.8 26.1–29.0 (overweight) 6.7 1.7 1.0 1.6 3.3 18.3 >29.0 (obese) 9.6 0.9 2.0 1.2 2.3 28.8 Any history of weight cycling (loss and regain of ≥6.8 kg)* No 66.0 70.8 77.1 80.2 67.8 46.3 Yes 34.0 29.2 22.9 19.8 32.2 53.7  Characteristic Entire population (n = 1,644) Adult net weight change (kg) ≤−2.5 (n = 120) −2.4 to 2.5 (n = 389) 2.6 to 4.9 (n = 257) 5.0 to 9.9 (n = 401) ≥10.0 (n = 477) Age (years) <20 0.7 0.7 2.3 0.4 0.0 0.2 20–34 70.6 77.5 74.4 71.2 71.1 64.8 35–39 23.4 20.8 19.8 23.0 23.9 26.8 ≥40 5.3 0.8 3.3 5.4 5.0 8.2 Race/ethnicity White 84.8 82.5 84.6 88.7 85.5 83.0 African American 1.9 0.8 2.3 0.0 2.0 2.7 Hispanic 2.9 3.3 2.3 2.0 3.2 3.4 Asian 7.4 10.0 9.5 6.2 6.5 6.5 Other 2.8 3.3 1.3 3.1 2.5 4.0 Missing 0.2 0.0 0.0 0.0 0.3 0.4 Nulliparous 67.7 65.0 67.9 71.2 70.6 63.9 Employed 82.6 80.0 81.0 84.4 83.8 82.6 ≤12 years of education 4.5 4.2 4.9 3.9 4.5 4.6 Household income ($) <30,000 4.3 5.8 6.4 2.3 3.5 4.0 30,001–69,999 24.2 24.2 18.0 22.2 24.7 29.8 ≥70,000 68.9 65.8 72.2 74.4 69.3 64.6 Missing 2.7 4.2 3.4 3.1 2.5 1.7 Regular physical activity in the year before pregnancy 91.9 91.7 92.3 91.4 94.3 89.7 Smoking history Never 71.9 70.8 75.1 77.2 71.0 67.4 Before pregnancy 21.7 20.0 19.2 18.4 23.7 24.0 At beginning of pregnancy 6.4 9.2 19.2 18.4 23.7 24.0 Prenatal vitamin use 97.4 97.5 97.2 98.4 98.3 96.2 First-degree familial diabetes 14.1 9.2 12.1 12.8 14.0 17.6 Pregravid body mass index (kg/m2) <19.8 (lean) 18.1 43.2 39.3 21.4 8.0 1.1 19.8–26.0 (normal) 65.6 54.2 57.6 75.9 86.4 51.8 26.1–29.0 (overweight) 6.7 1.7 1.0 1.6 3.3 18.3 >29.0 (obese) 9.6 0.9 2.0 1.2 2.3 28.8 Any history of weight cycling (loss and regain of ≥6.8 kg)* No 66.0 70.8 77.1 80.2 67.8 46.3 Yes 34.0 29.2 22.9 19.8 32.2 53.7
*

Weight cycling information is missing for two women.

Weight cycling is associated with weight gain (19). If weight gain is both an antecedent of weight cycling and an independent risk factor of gestational diabetes, control for net adult weight change is appropriate when examining the relation between weight cycling and the disorder. However, if weight gain results from weight cycling, controlling for net weight change may spuriously attenuate the estimated association with weight cycling by removing the influence of a causal mechanism. In consideration of the complex relations between weight gain and cycling, we examined the association between weight cycling and gestational diabetes risk without adjustment for adult weight change. We also examined the association within the subgroup of 477 women who gained 10 kg or more between age 18 years and the study pregnancy.

## RESULTS

Members of the analytical population were generally White, married, and affluent women (table 1). Slightly over half of the participants gained at least 5 kg between age 18 years and the study pregnancy, and 29 percent gained 10 kg or more. Fewer than 10 percent lost more than 2.5 kg during the same period. Adult net weight change was positively associated with maternal age. The proportion of Asian women was highest within the group of women who lost more than 2.5 kg in adulthood. Pregravid obesity and history of weight cycling were most prevalent among women who gained 10 kg or more during adulthood. Other characteristics were not strongly related to adult net weight change.

After adjustment for confounding maternal characteristics (age, race/ethnicity, and years of education), gestational diabetes risk was elevated among women who were obese at age 18 years compared with women of normal weight (relative risk (RR) = 4.53, 95 percent confidence interval (CI): 1.25, 16.43) (table 2). Risk was also higher among women who were lean at age 18 years (RR = 1.79, 95 percent CI: 1.01, 2.84). After adjustment for maternal characteristics, pregravid body mass index was positively associated with gestational diabetes risk (ptrend = 0.001): risk was lower among lean women (RR = 0.34, 95 percent CI: 0.12, 0.95) and higher among obese women (RR = 3.25, 95 percent CI: 1.85, 5.71) compared with their normal-weight counterparts. Overweight women were not at increased risk (RR = 0.74, 95 percent CI: 0.23, 2.40).

TABLE 2.

Relative risks and 95% confidence intervals of gestational diabetes mellitus in relation to weight and height characteristics, Seattle and Tacoma, Washington, 1996–2002

 Exposure Cases Total no. Unadjusted relative risk 95% confidence interval Adjusted relative risk 95% confidence interval No. % Body mass index at age 18 years (kg/m2)* <19.8 (lean) 38 5.0 762 1.71 1.03, 2.86 1.79 1.01, 2.84 19.8–26.0 (normal) 24 2.9 825 1.00 1.00 26.1–29.0 (overweight) 2 6.5 31 2.22 0.52, 9.38 2.63 0.62, 11.37 >29.0 (obese) 3 11.5 26 3.97 1.19, 13.17 4.53 1.25, 16.43 ptrend 0.61 0.72 Pregravid body mass index (kg/m2)* <19.8 (lean) 4 1.4 296 0.36 0.13, 0.99 0.34 0.12, 0.95 19.8–26.0 (normal) 41 3.8 1,078 1.00 1.00 26.1–29.0 (overweight) 3 2.8 109 0.72 0.22, 2.34 0.74 0.23, 2.40 >29.0 (obese) 19 11.8 161 3.10 1.80, 5.35 3.25 1.85, 5.71 ptrend 0.001 0.001 Adult height (cm)† ≤160 28 7.3 381 1.00 1.00 161–165 18 4.0 455 0.54 0.30, 0.97 0.70 0.38, 1.28 166–170 14 3.2 442 0.43 0.23, 0.82 0.63 0.32, 1.23 >170 7 1.9 366 0.26 0.11, 0.59 0.40 0.17, 0.95 ptrend <0.001 0.03 Net weight change from age 18 years to just before pregnancy (kg)‡ ≤−2.5 0 0.0 120 −2.4 to 2.5 9 2.3 389 1.00 1.00 2.6 to 4.9 8 3.1 257 1.35 0.52, 3.49 1.45 0.53, 3.96 5.0 to 9.9 17 4.2 401 1.83 0.82, 4.11 2.19 0.96, 5.05 ≥10.0 33 7.0 477 2.99 1.43, 6.25 3.43 1.60, 7.37 ptrend <0.001 <0.001 Any history of weight cycling (loss and regain of ≥6.8 kg)§ No 39 3.6 1,083 1.00 1.00 Yes 28 5.0 559 1.39 0.86, 2.26 0.93 0.53, 1.66 No. of weight cycles§ 0 39 3.6 1,083 1.00 1.00 1 8 2.9 272 0.82 0.38, 1.75 0.70 0.32, 1.53 2 8 5.6 143 1.55 0.73, 3.32 1.15 0.50, 2.65 ≥3 12 8.3 144 2.31 1.21, 4.42 1.23 0.56, 2.73 ptrend 0.01 0.69
 Exposure Cases Total no. Unadjusted relative risk 95% confidence interval Adjusted relative risk 95% confidence interval No. % Body mass index at age 18 years (kg/m2)* <19.8 (lean) 38 5.0 762 1.71 1.03, 2.86 1.79 1.01, 2.84 19.8–26.0 (normal) 24 2.9 825 1.00 1.00 26.1–29.0 (overweight) 2 6.5 31 2.22 0.52, 9.38 2.63 0.62, 11.37 >29.0 (obese) 3 11.5 26 3.97 1.19, 13.17 4.53 1.25, 16.43 ptrend 0.61 0.72 Pregravid body mass index (kg/m2)* <19.8 (lean) 4 1.4 296 0.36 0.13, 0.99 0.34 0.12, 0.95 19.8–26.0 (normal) 41 3.8 1,078 1.00 1.00 26.1–29.0 (overweight) 3 2.8 109 0.72 0.22, 2.34 0.74 0.23, 2.40 >29.0 (obese) 19 11.8 161 3.10 1.80, 5.35 3.25 1.85, 5.71 ptrend 0.001 0.001 Adult height (cm)† ≤160 28 7.3 381 1.00 1.00 161–165 18 4.0 455 0.54 0.30, 0.97 0.70 0.38, 1.28 166–170 14 3.2 442 0.43 0.23, 0.82 0.63 0.32, 1.23 >170 7 1.9 366 0.26 0.11, 0.59 0.40 0.17, 0.95 ptrend <0.001 0.03 Net weight change from age 18 years to just before pregnancy (kg)‡ ≤−2.5 0 0.0 120 −2.4 to 2.5 9 2.3 389 1.00 1.00 2.6 to 4.9 8 3.1 257 1.35 0.52, 3.49 1.45 0.53, 3.96 5.0 to 9.9 17 4.2 401 1.83 0.82, 4.11 2.19 0.96, 5.05 ≥10.0 33 7.0 477 2.99 1.43, 6.25 3.43 1.60, 7.37 ptrend <0.001 <0.001 Any history of weight cycling (loss and regain of ≥6.8 kg)§ No 39 3.6 1,083 1.00 1.00 Yes 28 5.0 559 1.39 0.86, 2.26 0.93 0.53, 1.66 No. of weight cycles§ 0 39 3.6 1,083 1.00 1.00 1 8 2.9 272 0.82 0.38, 1.75 0.70 0.32, 1.53 2 8 5.6 143 1.55 0.73, 3.32 1.15 0.50, 2.65 ≥3 12 8.3 144 2.31 1.21, 4.42 1.23 0.56, 2.73 ptrend 0.01 0.69
*

Adjusted model controls for age, race/ethnicity, and ≥12 years of education.

Adjusted model controls for age, race/ethnicity, education, and pregravid body mass index.

Adjusted model controls for age, race/ethnicity, parity, education, height, and categorical body mass index at age 18 years.

§

Weight cycling information is missing for two women. Adjusted model controls for age, race/ethnicity, parity, education, categorical body mass index at age 18 years, and continuous adult net weight change. Relative risks without adjustment for adult weight change are reported in the text.

Height was inversely associated with gestational diabetes risk after adjustment for maternal characteristics and pregravid body mass index (ptrend = 0.03). For instance, risk among women taller than 170 cm was approximately 60 percent lower than among women 160 cm or shorter (RR = 0.40, 95 percent CI: 0.17, 0.95).

Net adult weight change was positively associated with gestational diabetes risk after adjustment for maternal characteristics, height, and body mass index at age 18 years (ptrend < 0.001). Because none of the 120 women who lost more than 2.5 kg during adulthood also developed gestational diabetes, we could not estimate relative risks for this group. Risk was increased more than threefold among women who gained 10 kg or more in adulthood, compared with women with a weight change of 2.5 kg or less (RR = 3.43, 95 percent CI: 1.60, 7.37). After adjustment, a 5-kg gain between age 18 years and pregnancy was associated with a 20 percent increase in risk (95 percent CI: 1.14, 1.27). When adjusted for pregravid body mass index rather than body mass index at age 18 years, the net adult weight change remained associated with gestational diabetes risk, although the relative risks were somewhat attenuated. For instance, the relative risk associated with a weight change of 10 kg or more versus a change of less than or equal to 2.5 kg was 2.28 (95 percent CI: 0.95, 5.48). The relative risk for a 5-kg gain was 1.15 (95 percent CI: 1.05, 1.26).

Weight cycling was associated with gestational diabetes risk after adjustment for age, race/ethnicity, parity, education, and body mass index at age 18 years. The adjusted relative risk associated with any cycling history was 1.46 (95 percent CI: 0.87, 2.43). Adjusted relative risks for one, two, and three or more cycles were 0.85 (95 percent CI: 0.39, 1.83), 1.76 (95 percent CI: 0.81, 3.83), and 2.61 (95 percent CI: 1.24, 5.28) (ptrend = 0.008), respectively. However, weight cycling was not strongly associated after further adjustment for adult net weight change (table 2). For instance, the relative risk comparing three or more cycles with no cycles was 1.23 (95 percent CI: 0.56, 2.73).

Within the subgroup of women who had gained 10 kg or more in adulthood, any history of weight cycling was associated with a non-statistically significant increased risk after adjustment for maternal characteristics and body mass index at age 18 years (RR = 1.47, 95 percent CI: 0.70, 3.12) (table 3). The adjusted relative risk for three or more cycles, versus none, was 2.04 (95 percent CI: 0.83, 5.02). The ptrend across the number of cycles was 0.09. Small numbers hampered analyses of cycling and gestational diabetes risk within other subgroups classified by weight change.

TABLE 3.

Relative risks and 95% confidence intervals of gestational diabetes mellitus in relation to number of weight cycles among women who gained 10.0 kg or more between age 18 years and the study pregnancy, Seattle and Tacoma, Washington, 1996–2002

 Exposure Cases Total no. Unadjusted relative risk 95% confidence interval Adjusted relative risk 95% confidence interval No. % Any history of weight cycling (loss and regain of ≥6.8 kg)* No 12 5.5 220 1.00 1.00 Yes 21 8.2 255 1.29 0.96, 1.74 1.47 0.70, 3.12 No. of weight cycles† 0 12 5.5 220 1.00 1.00 1 5 5.2 96 0.95 0.34, 2.71 0.92 0.32, 2.69 2 6 8.6 70 1.57 0.59, 4.19 1.62 0.54, 4.32 ≥3 10 11.2 89 2.06 0.89, 4.77 2.04 0.83, 5.02 ptrend 0.01 0.09
 Exposure Cases Total no. Unadjusted relative risk 95% confidence interval Adjusted relative risk 95% confidence interval No. % Any history of weight cycling (loss and regain of ≥6.8 kg)* No 12 5.5 220 1.00 1.00 Yes 21 8.2 255 1.29 0.96, 1.74 1.47 0.70, 3.12 No. of weight cycles† 0 12 5.5 220 1.00 1.00 1 5 5.2 96 0.95 0.34, 2.71 0.92 0.32, 2.69 2 6 8.6 70 1.57 0.59, 4.19 1.62 0.54, 4.32 ≥3 10 11.2 89 2.06 0.89, 4.77 2.04 0.83, 5.02 ptrend 0.01 0.09
*

Adjusted model controls for age, race/ethnicity, parity, ≥12 years of education, and categorical body mass index at age 18 years.

Weight cycling information is missing for two women.

Weight characteristics before pregnancy may influence weight gain during pregnancy, which in turn may influence gestational diabetes risk. However, the disease itself and caloric restriction used in managing it may affect weight gain during pregnancy (1). We repeated adult weight change and weight cycling analyses after adjusting for maternal weight gain in early pregnancy, which is less likely to be influenced by the disease than weight gain until delivery. We measured early pregnancy weight gain as the difference between the weight recorded between 18 and 22 weeks of gestation (abstracted from medical records) and pregravid weight. These data were available for 97 percent of participants. Adjustment for early pregnancy weight gain and weeks of gestation at weight measurement did not substantially change relative risks for adult weight change and weight cycling. For instance, after adjustment for maternal characteristics, body mass index at 18 years, and early pregnancy weight gain, a 5-kg adult weight increase was associated with a 30 percent increase in risk (95 percent CI: 1.20, 1.41).

## DISCUSSION

In this study population, gestational diabetes risk was inversely related to height and directly related to pregravid body mass index. The relation with body mass index at age 18 years was J shaped, with higher risk among lean and obese women. Adult net weight gain was associated with increased risk independently of body mass index at age 18 years. The association was also independent of pregravid body mass index. Weight cycling was not substantially associated with risk after adjustment for adult weight change, although there was a nonsignificant association within the subgroup of women who had gained 10 kg or more during adulthood. The risks related to adult weight change and cycling were not explained by weight gain during early pregnancy.

These results should be interpreted in the context of the study limitations. First, height and weight characteristics were self-reported. In a previous study, although young women's self-reported height and weight correlated well with measured values, women slightly underreported their weight at age 18 years (20). Such misclassification, if present in the current study, is unlikely to be related to future gestational diabetes risk and would thus likely bias relative risk estimates toward the null. Second, because of small numbers, we were unable to estimate the association with adult net weight loss. We cannot infer from the risk elevations associated with adult weight gain that gestational diabetes risk is reduced after adult weight loss. Third, the interview assessment did not explicitly distinguish between intentional and unintentional weight cycling. Unintentional cycling is associated with poor health status and unhealthy behaviors (21). If women who had experienced unintentional weight cycles reported them as intentional, and if unintentional cycling is associated with increased gestational diabetes mellitus risk, the weight cycle relative risk estimates may be biased away from the null. However, the prevalence of past unintentional weight loss in this generally young and healthy population is likely to be low. Fourth, study participants were generally White, well educated, and affluent. To the extent that the characteristics may influence the relations we examined, the generalizability of these findings may be limited.

The associations with pregravid body mass index and height generally corroborate those previously reported in both pattern and strength (7, 8, 22–24). Solomon et al. (7) analyzed data from 14,613 members of the Nurses' Health Study II, aged 25–42 years, who did not have gestational diabetes or nongestational diabetes at enrollment and who delivered a singleton livebirth within the following 5 years. Women who were overweight at enrollment (body mass index: 25–29.9 kg/m2) had approximately a twofold increased risk of self-reported gestational diabetes (adjusted RR = 2.1, 95 percent CI: 1.6, 2.7) compared with lean women (body mass index: <20.0 kg/m2). The corresponding relative risk for obese women (body mass index: ≥30 kg/m2) was 2.9 (95 percent CI: 2.2, 3.9). In contrast, we did not observe an increased risk associated with overweight. The difference in body mass index categorizations may explain the disagreement in findings. Compared with the report of Solomon et al., our definition of overweight encompassed a slightly narrower range of body mass index values, and our referent group did not include very lean women. The inverse relation between height and gestational diabetes risk has been previously reported in ethnically diverse populations (22–24). For instance, among 9,005 Korean women, those in the lowest height quartile were at approximately twofold increased risk compared with the tallest quartile (odds ratio = 2.0, 95 percent CI: 1.4, 3.0) (22). This association is thought to reflect the influences of fetal and infant growth on both adult height and glucose tolerance (22, 24).

Although we observed increased risk among women who were lean at age 18 years, in the report of Solomon et al., leanness at age 18 years was not a risk factor (7). Risks were nearly equal among women with body mass indexes of 20–21.9 (RR = 0.9, 95 percent CI: 0.8, 1.1) and 22–24.9 kg/m2 (RR = 0.9, 95 percent CI: 0.7, 1.2) versus less than 20 kg/m2. Differences in study population characteristics may account for these inconsistencies. For instance, the prevalence of body mass index less than 20 kg/m2 at age 18 years was higher in our study population (50 percent vs. 39 percent). Smoking in adolescence may have confounded our estimate, as we did not collect information on teenage smoking. However, control for any history of smoking did not meaningfully change the relative risks.

Adult weight gain was associated with increased gestational diabetes risk, and this association weakened but remained after adjustment for pregravid body mass index. The association with adult weight gain described here is similar to that detected among a subset of Nurses' Health Study II participants (7). For instance, the adjusted relative risk comparing those who gained 20 kg or more between age 18 years and study enrollment (average age: 31 years) with those with less than 5-kg weight change was 3.6 (95 percent CI: 2.7, 4.7). Adult weight loss of more than 5 kg was not associated with gestational diabetes (adjusted RR = 0.9, 95 percent CI: 0.6, 1.4). Two studies have also reported increased gestational diabetes risk associated with weight gain between pregnancies (25, 26).

This is the first study, to our knowledge, to examine weight cycling in relation to gestational diabetes risk. Studies that have examined weight cycling in relation to type two diabetes in women provide inconsistent results (12, 13). Although weight cycling was associated with subsequent risk of type two diabetes among young and middle-aged Nurses' Health Study II participants, adjustment for body mass index after the period in which weight cycling was ascertained attenuated the association almost entirely (13). Given that weight cycling is a risk factor for weight gain among women (19), controlling for weight change during the period in which cycling is measured, or body mass index at the end of that period, may result in overadjusted relative risk estimates, that is, those that do not reflect the influence of a causal mechanism of weight cycling. The nonsignificant associations shown here among women who gained more than 10 kg in adulthood are not conclusive, but they suggest that avoidance of weight cycling may reduce risk, even if weight gain is not entirely prevented. These associations should be confirmed in other study populations. Diet and exercise choices that cause weight cycling may also influence gestational diabetes risk. Women classified as weight cyclers were more inclined than noncyclers in one study to change dietary habits rather than engage in frequent exercise for weight control (19). Prepregnancy exercise appears to reduce gestational diabetes risk independently of weight (27).

In summary, this report corroborates others by providing evidence that efforts to reduce overweight, obesity, and adult weight gain among young women may also reduce gestational diabetes risk. These results also motivate further examination of weight cycling as an independent risk factor for gestational diabetes.

### Abbreviations

Abbreviations
• CI

confidence interval

• RR

relative risk

Conflict of interest: none declared.

## References

1.
American Diabetes Association
Gestational diabetes mellitus
Diabetes Care
,
1999
, vol.
22

suppl
(pg.
S74
-
6
)
2.
O'Sullivan
JB
Camerini Davalos
RA
Cole
HS
Long term follow-up of gestational diabetes
Early diabetes
,
1984
New York, NY
(pg.
1009
-
27
)
3.
Jovanovic-Peterson
L
Peterson
CM
Is exercise safe or useful for gestational diabetic women?
Diabetes
,
1991
, vol.
40
(pg.
179
-
81
)
4.
Suhonen
L
Teramo
K
Hypertensive and preeclamptic women with gestational glucose intolerance
Acta Obstet Gynecol Scand
,
1993
, vol.
72
(pg.
269
-
72
)
5.
Foster-Powell
KA
Cheung
N
Recurrence of gestational diabetes
Aust N Z J Obstet Gynaecol
,
1998
, vol.
38
(pg.
384
-
7
)
6.
Damm
P
Gestational diabetes mellitus and subsequent development of overt diabetes mellitus
Dan Med Bull
,
1998
, vol.
45
(pg.
495
-
509
)
7.
Solomon
CG
Willet
WC
Carey
VJ
, et al.  .
A prospective study of pregravid determinants of gestational diabetes mellitus
JAMA
,
1997
, vol.
278
(pg.
1078
-
83
)
8.
Berkowitz
GS
Lapinski
RH
Wein
R
, et al.  .
Race/ethnicity and other risk factors for gestational diabetes
Am J Epidemiol
,
1992
, vol.
135
(pg.
965
-
73
)
9.
American College of Obstetricians and Gynecologists
ACOG committee opinion: obesity in pregnancy
Obstet Gynecol
,
2005
, vol.
106
(pg.
671
-
5
)
10.
Hedley
AA
Ogden
CL
Johnson
CL
, et al.  .
Prevalence of overweight and obesity among US children, adolescents, and adults, 1999 –2002
JAMA
,
2004
, vol.
291
(pg.
2847
-
50
)
11.
Prentice
AM
The emerging epidemic of obesity in developing countries
Int J Epidemiol
,
2006
, vol.
35
(pg.
93
-
9
)
12.
Morris
RD
Rimm
AA
Long-term weight fluctuation and non-insulin-dependent diabetes mellitus in white women
Ann Epidemiol
,
1992
, vol.
2
(pg.
657
-
64
)
13.
Field
AE
Manson
JE
Laird
N
, et al.  .
Weight cycling and the risk of developing type 2 diabetes among adult women in the United States
Obes Res
,
2004
, vol.
12
(pg.
267
-
74
)
14.
Report of the Expert Committee on the Diagnosis and Classification of Diabetes Mellitus
Diabetes Care
,
1997
, vol.
20
(pg.
1183
-
7
)
15.
Committee on Nutritional Status during Pregnancy and Lactation, Institute of Medicine
Nutrition during pregnancy: part I: weight gain, part II: nutrient supplements
,
1990
Washington, DC
Institute of Medicine
16.
Willett
WC
Manson
JE
Stampfer
MJ
, et al.  .
Weight, weight change, and coronary heart disease in women. Risk within the “normal” weight range
JAMA
,
1995
, vol.
273
(pg.
461
-
5
)
17.
Huang
Z
Willett
WC
Manson
JE
, et al.  .
Body weight, weight change, and risk of hypertension in women
Ann Intern Med
,
1998
, vol.
12
(pg.
81
-
7
)
18.
Hardin
JW
Hilbe
J
Generalized linear models and extensions
2001
College Station, TX
Stata Press
19.
Field
AE
Manson
JE
Taylor
CB
, et al.  .
Association of weight change, weight control practices, and weight cycling among women in the Nurses' Health Study II
Int J Obes Relat Metab Disord
,
2004
, vol.
28
(pg.
1134
-
42
)
20.
Troy
LM
Hunter
DJ
Manson
JE
, et al.  .
The validity of recalled weight among younger women
Int J Obes Relat Metab Disord
,
1995
, vol.
19
(pg.
570
-
2
)
21.
Meltzer
AA
Everhart
JE
Unintentional weight loss in the United States
Am J Epidemiol
,
1995
, vol.
142
(pg.
1039
-
46
)
22.
Jang
HC
Min
HK
Lee
HK
, et al.  .
Short stature in Korean women: a contribution to the multifactorial predisposition to gestational diabetes mellitus
Diabetologia
,
1998
, vol.
41
(pg.
778
-
83
)
23.
Kousta
E
Lawrence
NJ
Penny
A
, et al.  .
Women with a history of gestational diabetes of European and South Asian origin are shorter than women with normal glucose tolerance in pregnancy
Diabet Med
,
2000
, vol.
17
(pg.
792
-
7
)
24.
Anastasiou
E
Alevizaki
M
Grigorakis
SJ
, et al.  .
Decreased stature in gestational diabetes mellitus
Diabetologia
,
1998
, vol.
41
(pg.
997
-
1001
)
25.
Glazer
NL
Hendrickson
AF
Schellenbaum
GD
, et al.  .
Weight change and the risk of gestational diabetes in obese women
Epidemiology
,
2004
, vol.
15
(pg.
733
-
7
)
26.
Pole
JD
Dodds
LA
Maternal outcomes associated with weight change between pregnancies
Can J Public Health
,
1999
, vol.
90
(pg.
233
-
6
)
27.
Dempsey
JC
Sorensen
TK
Williams
MA
, et al.  .
Prospective study of gestational diabetes mellitus risk in relation to maternal recreational physical activity before and during pregnancy
Am J Epidemiol
,
2004
, vol.
159
(pg.
663
-
70
)